new queue structure

[unix-history] / usr / src / sys / ufs / ffs / ffs_inode.c

diff --git a/usr/src/sys/ufs/ffs/ffs_inode.c b/usr/src/sys/ufs/ffs/ffs_inode.c
index fb3a4b4..91a5ea7 100644 (file)
--- a/usr/src/sys/ufs/ffs/ffs_inode.c
+++ b/usr/src/sys/ufs/ffs/ffs_inode.c
@@ -1,351 +1,104 @@
 /*
 /*

- * Copyright (c) 1982, 1986, 1989 Regents of the University of California.
- * All rights reserved.
+ * Copyright (c) 1982, 1986, 1989, 1993
+ *     The Regents of the University of California.  All rights reserved.

  *
  * %sccs.include.redist.c%
  *
  *
  * %sccs.include.redist.c%
  *

- *     @(#)ffs_inode.c 7.33 (Berkeley) %G%
+ *     @(#)ffs_inode.c 8.5 (Berkeley) %G%

  */
 
  */
 

-#include "param.h"
-#include "systm.h"
-#include "mount.h"
-#include "user.h"
-#include "proc.h"
-#include "file.h"
-#include "buf.h"
-#include "cmap.h"
-#include "vnode.h"
-#include "../ufs/quota.h"
-#include "../ufs/inode.h"
-#include "../ufs/fs.h"
-#include "../ufs/ufsmount.h"
-#include "kernel.h"
-#include "malloc.h"
+#include <sys/param.h>
+#include <sys/systm.h>
+#include <sys/mount.h>
+#include <sys/proc.h>
+#include <sys/file.h>
+#include <sys/buf.h>
+#include <sys/vnode.h>
+#include <sys/kernel.h>
+#include <sys/malloc.h>
+#include <sys/trace.h>
+#include <sys/resourcevar.h>

 
 

-#define        INOHSZ  512
-#if    ((INOHSZ&(INOHSZ-1)) == 0)
-#define        INOHASH(dev,ino)        (((dev)+(ino))&(INOHSZ-1))
-#else
-#define        INOHASH(dev,ino)        (((unsigned)((dev)+(ino)))%INOHSZ)
-#endif
+#include <vm/vm.h>

 
 

-union ihead {
-       union  ihead *ih_head[2];
-       struct inode *ih_chain[2];
-} ihead[INOHSZ];
+#include <ufs/ufs/quota.h>
+#include <ufs/ufs/inode.h>
+#include <ufs/ufs/ufsmount.h>
+#include <ufs/ufs/ufs_extern.h>

 
 

-int prtactive; /* 1 => print out reclaim of active vnodes */
+#include <ufs/ffs/fs.h>
+#include <ufs/ffs/ffs_extern.h>

 
 

-/*
- * Initialize hash links for inodes.
- */
-ufs_init()
-{
-       register int i;
-       register union ihead *ih = ihead;
+static int ffs_indirtrunc __P((struct inode *, daddr_t, daddr_t, daddr_t, int,
+           long *));

 
 

-#ifndef lint
-       if (VN_MAXPRIVATE < sizeof(struct inode))
-               panic("ihinit: too small");
-#endif /* not lint */
-       for (i = INOHSZ; --i >= 0; ih++) {
-               ih->ih_head[0] = ih;
-               ih->ih_head[1] = ih;
-       }
-#ifdef QUOTA
-       dqinit();
-#endif /* QUOTA */
-}
-
-/*
- * Look up an vnode/inode by device,inumber.
- * If it is in core (in the inode structure),
- * honor the locking protocol.
- * If it is not in core, read it in from the
- * specified device.
- * Callers must check for mount points!!
- * In all cases, a pointer to a locked
- * inode structure is returned.
- */
-iget(xp, ino, ipp)
-       struct inode *xp;
-       ino_t ino;
-       struct inode **ipp;
+int
+ffs_init()

 {
 {

-       dev_t dev = xp->i_dev;
-       struct mount *mntp = ITOV(xp)->v_mount;
-       register struct fs *fs = VFSTOUFS(mntp)->um_fs;
-       extern struct vnodeops ufs_vnodeops, spec_inodeops;
-       register struct inode *ip, *iq;
-       register struct vnode *vp;
-       struct vnode *nvp;
-       struct buf *bp;
-       struct dinode *dp;
-       union ihead *ih;
-       int i, error;
-
-       ih = &ihead[INOHASH(dev, ino)];
-loop:
-       for (ip = ih->ih_chain[0]; ip != (struct inode *)ih; ip = ip->i_forw) {
-               if (ino != ip->i_number || dev != ip->i_dev)
-                       continue;
-               if ((ip->i_flag&ILOCKED) != 0) {
-                       ip->i_flag |= IWANT;
-                       sleep((caddr_t)ip, PINOD);
-                       goto loop;
-               }
-               if (vget(ITOV(ip)))
-                       goto loop;
-               *ipp = ip;
-               return(0);
-       }
-       /*
-        * Allocate a new inode.
-        */
-       if (error = getnewvnode(VT_UFS, mntp, &ufs_vnodeops, &nvp)) {
-               *ipp = 0;
-               return (error);
-       }
-       ip = VTOI(nvp);
-       ip->i_vnode = nvp;
-       ip->i_flag = 0;
-       ip->i_devvp = 0;
-       ip->i_mode = 0;
-       ip->i_diroff = 0;
-#ifdef QUOTA
-       for (i = 0; i < MAXQUOTAS; i++)
-               ip->i_dquot[i] = NODQUOT;
-#endif
-       /*
-        * Put it onto its hash chain and lock it so that other requests for
-        * this inode will block if they arrive while we are sleeping waiting
-        * for old data structures to be purged or for the contents of the
-        * disk portion of this inode to be read.
-        */
-       ip->i_dev = dev;
-       ip->i_number = ino;
-       insque(ip, ih);
-       ILOCK(ip);
-       /*
-        * Read in the disk contents for the inode.
-        */
-       if (error = bread(VFSTOUFS(mntp)->um_devvp, fsbtodb(fs, itod(fs, ino)),
-           (int)fs->fs_bsize, NOCRED, &bp)) {
-               /*
-                * The inode does not contain anything useful, so it would
-                * be misleading to leave it on its hash chain.
-                * Iput() will take care of putting it back on the free list.
-                */
-               remque(ip);
-               ip->i_forw = ip;
-               ip->i_back = ip;
-               /*
-                * Unlock and discard unneeded inode.
-                */
-               iput(ip);
-               brelse(bp);
-               *ipp = 0;
-               return (error);
-       }
-       dp = bp->b_un.b_dino;
-       dp += itoo(fs, ino);
-       ip->i_din = *dp;
-       brelse(bp);
-       /*
-        * Initialize the associated vnode
-        */
-       vp = ITOV(ip);
-       vp->v_type = IFTOVT(ip->i_mode);
-       if (vp->v_type == VFIFO) {
-#ifdef FIFO
-               extern struct vnodeops fifo_inodeops;
-               vp->v_op = &fifo_inodeops;
-#else
-               iput(ip);
-               *ipp = 0;
-               return (EOPNOTSUPP);
-#endif /* FIFO */
-       }
-       if (vp->v_type == VCHR || vp->v_type == VBLK) {
-               vp->v_op = &spec_inodeops;
-               if (nvp = checkalias(vp, ip->i_rdev, mntp)) {
-                       /*
-                        * Reinitialize aliased inode.
-                        */
-                       vp = nvp;
-                       iq = VTOI(vp);
-                       iq->i_vnode = vp;
-                       iq->i_flag = 0;
-                       ILOCK(iq);
-                       iq->i_din = ip->i_din;
-                       iq->i_dev = dev;
-                       iq->i_number = ino;
-                       insque(iq, ih);
-                       /*
-                        * Discard unneeded vnode
-                        */
-                       ip->i_mode = 0;
-                       iput(ip);
-                       ip = iq;
-               }
-       }
-       if (ino == ROOTINO)
-               vp->v_flag |= VROOT;
-       /*
-        * Finish inode initialization.
-        */
-       ip->i_fs = fs;
-       ip->i_devvp = VFSTOUFS(mntp)->um_devvp;
-       VREF(ip->i_devvp);
-       /*
-        * Set up a generation number for this inode if it does not
-        * already have one. This should only happen on old filesystems.
-        */
-       if (ip->i_gen == 0) {
-               if (++nextgennumber < (u_long)time.tv_sec)
-                       nextgennumber = time.tv_sec;
-               ip->i_gen = nextgennumber;
-               if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
-                       ip->i_flag |= IMOD;
-       }
-       *ipp = ip;
-       return (0);
+       return (ufs_init());

 }
 
 /*
 }
 
 /*

- * Unlock and decrement the reference count of an inode structure.
+ * Update the access, modified, and inode change times as specified by the
+ * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is
+ * used to specify that the inode needs to be updated but that the times have
+ * already been set. The access and modified times are taken from the second
+ * and third parameters; the inode change time is always taken from the current
+ * time. If waitfor is set, then wait for the disk write of the inode to
+ * complete.

  */
  */

-iput(ip)
-       register struct inode *ip;
+int
+ffs_update(ap)
+       struct vop_update_args /* {
+               struct vnode *a_vp;
+               struct timeval *a_access;
+               struct timeval *a_modify;
+               int a_waitfor;
+       } */ *ap;

 {
 {

+       register struct fs *fs;
+       struct buf *bp;
+       struct inode *ip;
+       register struct fs *fs;

 
 

-       if ((ip->i_flag & ILOCKED) == 0)
-               panic("iput");
-       IUNLOCK(ip);
-       vrele(ITOV(ip));
-}
-
-/*
- * Last reference to an inode, write the inode out and if necessary,
- * truncate and deallocate the file.
- */
-ufs_inactive(vp)
-       struct vnode *vp;
-{
-       register struct inode *ip = VTOI(vp);
-       int mode, error = 0;
-
-       if (prtactive && vp->v_usecount != 0)
-               vprint("ufs_inactive: pushing active", vp);
-       /*
-        * Get rid of inodes related to stale file handles.
-        */
-       if (ip->i_mode == 0) {
-               if ((vp->v_flag & VXLOCK) == 0)
-                       vgone(vp);
+       ip = VTOI(ap->a_vp);
+       if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY) {
+               ip->i_flag &=
+                   ~(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE);

                return (0);
        }
                return (0);
        }

-       ILOCK(ip);
-       if (ip->i_nlink <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
-#ifdef QUOTA
-               if (!getinoquota(ip))
-                       (void) chkiq(ip, -1, NOCRED, 0);
-#endif
-               error = itrunc(ip, (u_long)0, 0);
-               mode = ip->i_mode;
-               ip->i_mode = 0;
-               ip->i_flag |= IUPD|ICHG;
-               ifree(ip, ip->i_number, mode);
+       if ((ip->i_flag &
+           (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0)
+               return (0);
+       if (ip->i_flag & IN_ACCESS)
+               ip->i_atime.ts_sec = ap->a_access->tv_sec;
+       if (ip->i_flag & IN_UPDATE) {
+               ip->i_mtime.ts_sec = ap->a_modify->tv_sec;
+               ip->i_modrev++;

        }
        }

-       IUPDAT(ip, &time, &time, 0);
-       IUNLOCK(ip);
-       ip->i_flag = 0;
-       /*
-        * If we are done with the inode, reclaim it
-        * so that it can be reused immediately.
-        */
-       if (vp->v_usecount == 0 && ip->i_mode == 0)
-               vgone(vp);
-       return (error);
-}
-
-/*
- * Reclaim an inode so that it can be used for other purposes.
- */
-ufs_reclaim(vp)
-       register struct vnode *vp;
-{
-       register struct inode *ip = VTOI(vp);
-       int i;
-
-       if (prtactive && vp->v_usecount != 0)
-               vprint("ufs_reclaim: pushing active", vp);
-       /*
-        * Remove the inode from its hash chain.
-        */
-       remque(ip);
-       ip->i_forw = ip;
-       ip->i_back = ip;
+       if (ip->i_flag & IN_CHANGE)
+               ip->i_ctime.ts_sec = time.tv_sec;
+       ip->i_flag &= ~(IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE);
+       fs = ip->i_fs;

        /*
        /*

-        * Purge old data structures associated with the inode.
+        * Ensure that uid and gid are correct. This is a temporary
+        * fix until fsck has been changed to do the update.

         */
         */

-       cache_purge(vp);
-       if (ip->i_devvp) {
-               vrele(ip->i_devvp);
-               ip->i_devvp = 0;
-       }
-#ifdef QUOTA
-       for (i = 0; i < MAXQUOTAS; i++) {
-               if (ip->i_dquot[i] != NODQUOT) {
-                       dqrele(vp, ip->i_dquot[i]);
-                       ip->i_dquot[i] = NODQUOT;
-               }
-       }
-#endif
-       ip->i_flag = 0;
-       return (0);
-}
-
-/*
- * Check accessed and update flags on an inode structure.
- * If any is on, update the inode with the current time.
- * If waitfor is given, then must ensure I/O order,
- * so wait for write to complete.
- */
-iupdat(ip, ta, tm, waitfor)
-       register struct inode *ip;
-       struct timeval *ta, *tm;
-       int waitfor;
-{
-       struct buf *bp;
-       struct vnode *vp = ITOV(ip);
-       struct dinode *dp;
-       register struct fs *fs;
-
-       fs = ip->i_fs;
-       if ((ip->i_flag & (IUPD|IACC|ICHG|IMOD)) == 0)
-               return (0);
-       if (vp->v_mount->mnt_flag & MNT_RDONLY)
-               return (0);
-       error = bread(ip->i_devvp, fsbtodb(fs, itod(fs, ip->i_number)),
-               (int)fs->fs_bsize, NOCRED, &bp);
-       if (error) {
+       if (fs->fs_inodefmt < FS_44INODEFMT) {          /* XXX */
+               ip->i_din.di_ouid = ip->i_uid;          /* XXX */
+               ip->i_din.di_ogid = ip->i_gid;          /* XXX */
+       }                                               /* XXX */
+       if (error = bread(ip->i_devvp,
+           fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
+               (int)fs->fs_bsize, NOCRED, &bp)) {

                brelse(bp);
                return (error);
        }
                brelse(bp);
                return (error);
        }

-       if (ip->i_flag&IACC)
-               ip->i_atime = ta->tv_sec;
-       if (ip->i_flag&IUPD)
-               ip->i_mtime = tm->tv_sec;
-       if (ip->i_flag&ICHG)
-               ip->i_ctime = time.tv_sec;
-       ip->i_flag &= ~(IUPD|IACC|ICHG|IMOD);
-       dp = bp->b_un.b_dino + itoo(fs, ip->i_number);
-       *dp = ip->i_din;
-       if (waitfor) {
+       *((struct dinode *)bp->b_data +
+           ino_to_fsbo(fs, ip->i_number)) = ip->i_din;
+       if (ap->a_waitfor)

                return (bwrite(bp));
                return (bwrite(bp));

-       } else {
+       else {

                bdwrite(bp);
                return (0);
        }
                bdwrite(bp);
                return (0);
        }


@@ -355,88 +108,126 @@ iupdat(ip, ta, tm, waitfor)
 #define        DOUBLE  1       /* index of double indirect block */
 #define        TRIPLE  2       /* index of triple indirect block */
 /*
 #define        DOUBLE  1       /* index of double indirect block */
 #define        TRIPLE  2       /* index of triple indirect block */
 /*

- * Truncate the inode ip to at most length size.  Free affected disk
- * blocks -- the blocks of the file are removed in reverse order.
- *
- * NB: triple indirect blocks are untested.
+ * Truncate the inode oip to at most length size, freeing the
+ * disk blocks.

  */
  */

-itrunc(oip, length, flags)
-       register struct inode *oip;
-       u_long length;
-       int flags;
+ffs_truncate(ap)
+       struct vop_truncate_args /* {
+               struct vnode *a_vp;
+               off_t a_length;
+               int a_flags;
+               struct ucred *a_cred;
+               struct proc *a_p;
+       } */ *ap;

 {
 {

+       register struct vnode *ovp = ap->a_vp;

        register daddr_t lastblock;
        register daddr_t lastblock;

-       daddr_t bn, lbn, lastiblock[NIADDR];
+       register struct inode *oip;
+       daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
+       daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
+       off_t length = ap->a_length;

        register struct fs *fs;
        register struct fs *fs;

-       register struct inode *ip;

        struct buf *bp;
        struct buf *bp;

-       int offset, osize, size, level;
-       long count, nblocks, blocksreleased = 0;
+       int offset, size, level;
+       long count, nblocks, vflags, blocksreleased = 0;
+       struct timeval tv;

        register int i;
        int aflags, error, allerror;
        register int i;
        int aflags, error, allerror;

-       struct inode tip;
+       off_t osize;

 
 

-       if (oip->i_size <= length) {
-               oip->i_flag |= ICHG|IUPD;
-               error = iupdat(oip, &time, &time, 1);
-               return (error);
+       oip = VTOI(ovp);
+       tv = time;
+       if (ovp->v_type == VLNK &&
+           oip->i_size < ovp->v_mount->mnt_maxsymlinklen) {
+#ifdef DIAGNOSTIC
+               if (length != 0)
+                       panic("ffs_truncate: partial truncate of symlink");
+#endif
+               bzero((char *)&oip->i_shortlink, (u_int)oip->i_size);
+               oip->i_size = 0;
+               oip->i_flag |= IN_CHANGE | IN_UPDATE;
+               return (VOP_UPDATE(ovp, &tv, &tv, 1));

        }
        }

+       if (oip->i_size == length) {
+               oip->i_flag |= IN_CHANGE | IN_UPDATE;
+               return (VOP_UPDATE(ovp, &tv, &tv, 0));
+       }
+#ifdef QUOTA
+       if (error = getinoquota(oip))
+               return (error);
+#endif
+       vnode_pager_setsize(ovp, (u_long)length);
+       fs = oip->i_fs;
+       osize = oip->i_size;

        /*
        /*

-        * Calculate index into inode's block list of
-        * last direct and indirect blocks (if any)
-        * which we want to keep.  Lastblock is -1 when
-        * the file is truncated to 0.
+        * Lengthen the size of the file. We must ensure that the
+        * last byte of the file is allocated. Since the smallest
+        * value of oszie is 0, length will be at least 1.

         */
         */

-       fs = oip->i_fs;
-       lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
-       lastiblock[SINGLE] = lastblock - NDADDR;
-       lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
-       lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
-       nblocks = btodb(fs->fs_bsize);
+       if (osize < length) {
+               offset = blkoff(fs, length - 1);
+               lbn = lblkno(fs, length - 1);
+               aflags = B_CLRBUF;
+               if (ap->a_flags & IO_SYNC)
+                       aflags |= B_SYNC;
+               if (error = ffs_balloc(oip, lbn, offset + 1, ap->a_cred, &bp,
+                   aflags))
+                       return (error);
+               oip->i_size = length;
+               (void) vnode_pager_uncache(ovp);
+               if (aflags & IO_SYNC)
+                       bwrite(bp);
+               else
+                       bawrite(bp);
+               oip->i_flag |= IN_CHANGE | IN_UPDATE;
+               return (VOP_UPDATE(ovp, &tv, &tv, 1));
+       }

        /*
        /*

-        * Update the size of the file. If the file is not being
+        * Shorten the size of the file. If the file is not being

         * truncated to a block boundry, the contents of the
         * partial block following the end of the file must be
         * zero'ed in case it ever become accessable again because
         * of subsequent file growth.
         */
         * truncated to a block boundry, the contents of the
         * partial block following the end of the file must be
         * zero'ed in case it ever become accessable again because
         * of subsequent file growth.
         */

-       osize = oip->i_size;

        offset = blkoff(fs, length);
        if (offset == 0) {
                oip->i_size = length;
        } else {
                lbn = lblkno(fs, length);
                aflags = B_CLRBUF;
        offset = blkoff(fs, length);
        if (offset == 0) {
                oip->i_size = length;
        } else {
                lbn = lblkno(fs, length);
                aflags = B_CLRBUF;

-               if (flags & IO_SYNC)
+               if (ap->a_flags & IO_SYNC)

                        aflags |= B_SYNC;
                        aflags |= B_SYNC;

-#ifdef QUOTA
-               if (error = getinoquota(oip))
-                       return (error);
-#endif
-               if (error = balloc(oip, lbn, offset, &bp, aflags))
+               if (error = ffs_balloc(oip, lbn, offset, ap->a_cred, &bp,
+                   aflags))

                        return (error);
                oip->i_size = length;
                size = blksize(fs, oip, lbn);
                        return (error);
                oip->i_size = length;
                size = blksize(fs, oip, lbn);

-               bn = bp->b_blkno;
-               count = howmany(size, CLBYTES);
-                       munhash(oip->i_devvp, bn + i * CLBYTES / DEV_BSIZE);
-               bzero(bp->b_un.b_addr + offset, (unsigned)(size - offset));
-               brealloc(bp, size);
-               if (flags & IO_SYNC)
+               (void) vnode_pager_uncache(ovp);
+               bzero((char *)bp->b_data + offset, (u_int)(size - offset));
+               allocbuf(bp, size);
+               if (aflags & IO_SYNC)

                        bwrite(bp);
                else
                        bwrite(bp);
                else

-                       bdwrite(bp);
+                       bawrite(bp);

        }
        /*
        }
        /*

-        * Update file and block pointers
-        * on disk before we start freeing blocks.
-        * If we crash before free'ing blocks below,
-        * the blocks will be returned to the free list.
-        * lastiblock values are also normalized to -1
-        * for calls to indirtrunc below.
+        * Calculate index into inode's block list of
+        * last direct and indirect blocks (if any)
+        * which we want to keep.  Lastblock is -1 when
+        * the file is truncated to 0.

         */
         */

-       tip = *oip;
-       tip.i_size = osize;
+       lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
+       lastiblock[SINGLE] = lastblock - NDADDR;
+       lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
+       lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
+       nblocks = btodb(fs->fs_bsize);
+       /*
+        * Update file and block pointers on disk before we start freeing
+        * blocks.  If we crash before free'ing blocks below, the blocks
+        * will be returned to the free list.  lastiblock values are also
+        * normalized to -1 for calls to ffs_indirtrunc below.
+        */
+       bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof oldblks);

        for (level = TRIPLE; level >= SINGLE; level--)
                if (lastiblock[level] < 0) {
                        oip->i_ib[level] = 0;
        for (level = TRIPLE; level >= SINGLE; level--)
                if (lastiblock[level] < 0) {
                        oip->i_ib[level] = 0;


@@ -444,25 +235,38 @@ itrunc(oip, length, flags)
                }
        for (i = NDADDR - 1; i > lastblock; i--)
                oip->i_db[i] = 0;
                }
        for (i = NDADDR - 1; i > lastblock; i--)
                oip->i_db[i] = 0;

-       oip->i_flag |= ICHG|IUPD;
-       vinvalbuf(ITOV(oip), (length > 0));
-       allerror = iupdat(oip, &time, &time, MNT_WAIT);
+       oip->i_flag |= IN_CHANGE | IN_UPDATE;
+       if (error = VOP_UPDATE(ovp, &tv, &tv, MNT_WAIT))
+               allerror = error;
+       /*
+        * Having written the new inode to disk, save its new configuration
+        * and put back the old block pointers long enough to process them.
+        * Note that we save the new block configuration so we can check it
+        * when we are done.
+        */
+       bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof newblks);
+       bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof oldblks);
+       oip->i_size = osize;
+       vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA;
+       allerror = vinvalbuf(ovp, vflags, ap->a_cred, ap->a_p, 0, 0);

 
        /*
         * Indirect blocks first.
         */
 
        /*
         * Indirect blocks first.
         */

-       ip = &tip;
+       indir_lbn[SINGLE] = -NDADDR;
+       indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
+       indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;

        for (level = TRIPLE; level >= SINGLE; level--) {
        for (level = TRIPLE; level >= SINGLE; level--) {

-               bn = ip->i_ib[level];
+               bn = oip->i_ib[level];

                if (bn != 0) {
                if (bn != 0) {

-                       error = indirtrunc(ip, bn, lastiblock[level], level,
-                               &count);
+                       error = ffs_indirtrunc(oip, indir_lbn[level],
+                           fsbtodb(fs, bn), lastiblock[level], level, &count);

                        if (error)
                                allerror = error;
                        blocksreleased += count;
                        if (lastiblock[level] < 0) {
                        if (error)
                                allerror = error;
                        blocksreleased += count;
                        if (lastiblock[level] < 0) {

-                               ip->i_ib[level] = 0;
-                               blkfree(ip, bn, (off_t)fs->fs_bsize);
+                               oip->i_ib[level] = 0;
+                               ffs_blkfree(oip, bn, fs->fs_bsize);

                                blocksreleased += nblocks;
                        }
                }
                                blocksreleased += nblocks;
                        }
                }


@@ -474,14 +278,14 @@ itrunc(oip, length, flags)
         * All whole direct blocks or frags.
         */
        for (i = NDADDR - 1; i > lastblock; i--) {
         * All whole direct blocks or frags.
         */
        for (i = NDADDR - 1; i > lastblock; i--) {

-               register off_t bsize;
+               register long bsize;

 
 

-               bn = ip->i_db[i];
+               bn = oip->i_db[i];

                if (bn == 0)
                        continue;
                if (bn == 0)
                        continue;

-               ip->i_db[i] = 0;
-               bsize = (off_t)blksize(fs, ip, i);
-               blkfree(ip, bn, bsize);
+               oip->i_db[i] = 0;
+               bsize = blksize(fs, oip, i);
+               ffs_blkfree(oip, bn, bsize);

                blocksreleased += btodb(bsize);
        }
        if (lastblock < 0)
                blocksreleased += btodb(bsize);
        }
        if (lastblock < 0)


@@ -491,17 +295,17 @@ itrunc(oip, length, flags)
         * Finally, look for a change in size of the
         * last direct block; release any frags.
         */
         * Finally, look for a change in size of the
         * last direct block; release any frags.
         */

-       bn = ip->i_db[lastblock];
+       bn = oip->i_db[lastblock];

        if (bn != 0) {
        if (bn != 0) {

-               off_t oldspace, newspace;
+               long oldspace, newspace;

 
                /*
                 * Calculate amount of space we're giving
                 * back as old block size minus new block size.
                 */
 
                /*
                 * Calculate amount of space we're giving
                 * back as old block size minus new block size.
                 */

-               oldspace = blksize(fs, ip, lastblock);
-               ip->i_size = length;
-               newspace = blksize(fs, ip, lastblock);
+               oldspace = blksize(fs, oip, lastblock);
+               oip->i_size = length;
+               newspace = blksize(fs, oip, lastblock);

                if (newspace == 0)
                        panic("itrunc: newspace");
                if (oldspace - newspace > 0) {
                if (newspace == 0)
                        panic("itrunc: newspace");
                if (oldspace - newspace > 0) {


@@ -511,43 +315,50 @@ itrunc(oip, length, flags)
                         * required for the storage we're keeping.
                         */
                        bn += numfrags(fs, newspace);
                         * required for the storage we're keeping.
                         */
                        bn += numfrags(fs, newspace);

-                       blkfree(ip, bn, oldspace - newspace);
+                       ffs_blkfree(oip, bn, oldspace - newspace);

                        blocksreleased += btodb(oldspace - newspace);
                }
        }
 done:
                        blocksreleased += btodb(oldspace - newspace);
                }
        }
 done:

-/* BEGIN PARANOIA */
+#ifdef DIAGNOSTIC

        for (level = SINGLE; level <= TRIPLE; level++)
        for (level = SINGLE; level <= TRIPLE; level++)

-               if (ip->i_ib[level] != oip->i_ib[level])
+               if (newblks[NDADDR + level] != oip->i_ib[level])

                        panic("itrunc1");
        for (i = 0; i < NDADDR; i++)
                        panic("itrunc1");
        for (i = 0; i < NDADDR; i++)

-               if (ip->i_db[i] != oip->i_db[i])
+               if (newblks[i] != oip->i_db[i])

                        panic("itrunc2");
                        panic("itrunc2");

-/* END PARANOIA */
+       if (length == 0 &&
+           (ovp->v_dirtyblkhd.lh_first || ovp->v_cleanblkhd.lh_first))
+               panic("itrunc3");
+#endif /* DIAGNOSTIC */
+       /*
+        * Put back the real size.
+        */
+       oip->i_size = length;

        oip->i_blocks -= blocksreleased;
        if (oip->i_blocks < 0)                  /* sanity */
                oip->i_blocks = 0;
        oip->i_blocks -= blocksreleased;
        if (oip->i_blocks < 0)                  /* sanity */
                oip->i_blocks = 0;

-       oip->i_flag |= ICHG;
+       oip->i_flag |= IN_CHANGE;

 #ifdef QUOTA
 #ifdef QUOTA

-       if (!getinoquota(oip))
-               (void) chkdq(oip, -blocksreleased, NOCRED, 0);
+       (void) chkdq(oip, -blocksreleased, NOCRED, 0);

 #endif
        return (allerror);
 }
 
 /*
 #endif
        return (allerror);
 }
 
 /*

- * Release blocks associated with the inode ip and
- * stored in the indirect block bn.  Blocks are free'd
- * in LIFO order up to (but not including) lastbn.  If
- * level is greater than SINGLE, the block is an indirect
- * block and recursive calls to indirtrunc must be used to
- * cleanse other indirect blocks.
+ * Release blocks associated with the inode ip and stored in the indirect
+ * block bn.  Blocks are free'd in LIFO order up to (but not including)
+ * lastbn.  If level is greater than SINGLE, the block is an indirect block
+ * and recursive calls to indirtrunc must be used to cleanse other indirect
+ * blocks.

  *
  * NB: triple indirect blocks are untested.
  */
  *
  * NB: triple indirect blocks are untested.
  */

-indirtrunc(ip, bn, lastbn, level, countp)
+static int
+ffs_indirtrunc(ip, lbn, dbn, lastbn, level, countp)

        register struct inode *ip;
        register struct inode *ip;

-       daddr_t bn, lastbn;
+       daddr_t lbn, lastbn;
+       daddr_t dbn;

        int level;
        long *countp;
 {
        int level;
        long *countp;
 {


@@ -555,10 +366,11 @@ indirtrunc(ip, bn, lastbn, level, countp)
        struct buf *bp;
        register struct fs *fs = ip->i_fs;
        register daddr_t *bap;
        struct buf *bp;
        register struct fs *fs = ip->i_fs;
        register daddr_t *bap;

-       daddr_t *copy, nb, last;
+       struct vnode *vp;
+       daddr_t *copy, nb, nlbn, last;

        long blkcount, factor;
        int nblocks, blocksreleased = 0;
        long blkcount, factor;
        int nblocks, blocksreleased = 0;

-       int error, allerror = 0;
+       int error = 0, allerror = 0;

 
        /*
         * Calculate index in current block of last
 
        /*
         * Calculate index in current block of last


@@ -573,22 +385,39 @@ indirtrunc(ip, bn, lastbn, level, countp)
                last /= factor;
        nblocks = btodb(fs->fs_bsize);
        /*
                last /= factor;
        nblocks = btodb(fs->fs_bsize);
        /*

-        * Get buffer of block pointers, zero those 
-        * entries corresponding to blocks to be free'd,
-        * and update on disk copy first.
+        * Get buffer of block pointers, zero those entries corresponding
+        * to blocks to be free'd, and update on disk copy first.  Since
+        * double(triple) indirect before single(double) indirect, calls
+        * to bmap on these blocks will fail.  However, we already have
+        * the on disk address, so we have to set the b_blkno field
+        * explicitly instead of letting bread do everything for us.

         */
 #ifdef SECSIZE
        bp = bread(ip->i_dev, fsbtodb(fs, bn), (int)fs->fs_bsize,
            fs->fs_dbsize);
 #else SECSIZE
         */
 #ifdef SECSIZE
        bp = bread(ip->i_dev, fsbtodb(fs, bn), (int)fs->fs_bsize,
            fs->fs_dbsize);
 #else SECSIZE

-       error = bread(ip->i_devvp, fsbtodb(fs, bn), (int)fs->fs_bsize,
-               NOCRED, &bp);
+       vp = ITOV(ip);
+       bp = getblk(vp, lbn, (int)fs->fs_bsize, 0, 0);
+       if (bp->b_flags & (B_DONE | B_DELWRI)) {
+               /* Braces must be here in case trace evaluates to nothing. */
+               trace(TR_BREADHIT, pack(vp, fs->fs_bsize), lbn);
+       } else {
+               trace(TR_BREADMISS, pack(vp, fs->fs_bsize), lbn);
+               curproc->p_stats->p_ru.ru_inblock++;    /* pay for read */
+               bp->b_flags |= B_READ;
+               if (bp->b_bcount > bp->b_bufsize)
+                       panic("ffs_indirtrunc: bad buffer size");
+               bp->b_blkno = dbn;
+               VOP_STRATEGY(bp);
+               error = biowait(bp);
+       }

        if (error) {
                brelse(bp);
                *countp = 0;
                return (error);
        }
        if (error) {
                brelse(bp);
                *countp = 0;
                return (error);
        }

-       bap = bp->b_un.b_daddr;
+
+       bap = (daddr_t *)bp->b_data;

        MALLOC(copy, daddr_t *, fs->fs_bsize, M_TEMP, M_WAITOK);
        bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->fs_bsize);
        bzero((caddr_t)&bap[last + 1],
        MALLOC(copy, daddr_t *, fs->fs_bsize, M_TEMP, M_WAITOK);
        bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->fs_bsize);
        bzero((caddr_t)&bap[last + 1],


@@ -603,18 +432,18 @@ indirtrunc(ip, bn, lastbn, level, countp)
        /*
         * Recursively free totally unused blocks.
         */
        /*
         * Recursively free totally unused blocks.
         */

-       for (i = NINDIR(fs) - 1; i > last; i--) {
+       for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
+           i--, nlbn += factor) {

                nb = bap[i];
                if (nb == 0)
                        continue;
                if (level > SINGLE) {
                nb = bap[i];
                if (nb == 0)
                        continue;
                if (level > SINGLE) {

-                       error = indirtrunc(ip, nb, (daddr_t)-1, level - 1,
-                               &blkcount);
-                       if (error)
+                       if (error = ffs_indirtrunc(ip, nlbn,
+                           fsbtodb(fs, nb), (daddr_t)-1, level - 1, &blkcount))

                                allerror = error;
                        blocksreleased += blkcount;
                }
                                allerror = error;
                        blocksreleased += blkcount;
                }

-               blkfree(ip, nb, (off_t)fs->fs_bsize);
+               ffs_blkfree(ip, nb, fs->fs_bsize);

                blocksreleased += nblocks;
        }
 
                blocksreleased += nblocks;
        }
 


@@ -625,8 +454,8 @@ indirtrunc(ip, bn, lastbn, level, countp)
                last = lastbn % factor;
                nb = bap[i];
                if (nb != 0) {
                last = lastbn % factor;
                nb = bap[i];
                if (nb != 0) {

-                       error = indirtrunc(ip, nb, last, level - 1, &blkcount);
-                       if (error)
+                       if (error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
+                           last, level - 1, &blkcount))

                                allerror = error;
                        blocksreleased += blkcount;
                }
                                allerror = error;
                        blocksreleased += blkcount;
                }


@@ -635,41 +464,3 @@ indirtrunc(ip, bn, lastbn, level, countp)
        *countp = blocksreleased;
        return (allerror);
 }
        *countp = blocksreleased;
        return (allerror);
 }

-
-/*
- * Lock an inode. If its already locked, set the WANT bit and sleep.
- */
-ilock(ip)
-       register struct inode *ip;
-{
-
-       while (ip->i_flag & ILOCKED) {
-               ip->i_flag |= IWANT;
-               if (ip->i_spare0 == u.u_procp->p_pid)
-                       panic("locking against myself");
-               ip->i_spare1 = u.u_procp->p_pid;
-               (void) sleep((caddr_t)ip, PINOD);
-       }
-       ip->i_spare1 = 0;
-       ip->i_spare0 = u.u_procp->p_pid;
-       u.u_spare[0]++;
-       ip->i_flag |= ILOCKED;
-}
-
-/*
- * Unlock an inode.  If WANT bit is on, wakeup.
- */
-iunlock(ip)
-       register struct inode *ip;
-{
-
-       if ((ip->i_flag & ILOCKED) == 0)
-               vprint("iunlock: unlocked inode", ITOV(ip));
-       ip->i_spare0 = 0;
-       u.u_spare[0]--;
-       ip->i_flag &= ~ILOCKED;
-       if (ip->i_flag&IWANT) {
-               ip->i_flag &= ~IWANT;
-               wakeup((caddr_t)ip);
-       }
-}